AI handles
- Drafting prose
- Adapting tone for each client
- Personalising voice for each advisor
- Suggesting structure and emphasis
Investment communications - For RIAs and private banks
Our data - your view.
AI drafts. Code verifies. Your Investment Office decides.
RiffSharp pulls data, drafts candidate views, and surfaces analysis for approval. Advisors get the full briefing; clients get the same approved view in language tuned to them. Client data stays inside your perimeter.
Workflow automation, not AI hype Tokenised at your edge Tailored by investor profile x DISC Built for BaFin - FINMA - CSSF - SEC scrutiny
Mid-market investment committees cover thousands of names with two to four people. The bottleneck is not judgment - it is the legwork ahead of judgment: pulling press releases, parsing transcripts, comparing consensus, drafting structured research.
RiffSharp does that work. We pull from authoritative sources - SEC filings, FRED, market-data APIs, transcript and flow feeds - and produce a candidate house view with the analysis already done.
Your Investment Office reviews, modifies, and approves across six clear decisions: rating and target, conviction, evidence, counter-arguments, distribution, and framing for advisors and clients.
Once approved, RiffSharp delivers everywhere it needs to go. Advisors get the full briefing - talking points, context, evidence, counter-arguments, what to expect on the call. Clients get the same approved view, in language tuned to their investor profile and DISC personality type.
Judgment time goes where it earns the firm's fees. The legwork - and the personalisation - go where they scale.
Next-generation HNW clients expect digital-native engagement before they ever pick up the phone. Capgemini's most recent World Wealth Report puts the inheritor switch rate at over 80% within two years of wealth transfer. The decision is rarely a complaint - it is a quiet reassessment that begins long before the relationship ends.
On the other side of the bench, banker mobility is at a record high. McKinsey's most recent practitioner survey finds roughly four out of five departing private bankers cite outdated client-facing tooling as a primary or contributing reason. Not compensation. Not culture. The platform.
Firms that move first set the standard. Firms that wait adopt under pressure - on someone else's terms.
Roughly seven in ten bank AI projects fail. Every institution we talk to has scars - pilots stuck in the sandbox, models that hallucinated numbers, systems no auditor would sign, vendors that promised regulatory readiness and did not deliver.
RiffSharp was built differently. We are industrial-grade workflow automation, the way regulated systems have always been built. Determinism where it matters. Verification at every layer. Configurability per firm. A full audit trail. Nothing magical.
Language models do what they are genuinely good at: drafting prose, adapting tone. Every claim is anchored to a source. Every number a client ever sees came from the approved Fact Bundle, checked byte-for-byte before delivery. No language model performs an investment computation. No model decides what a client receives. Your Investment Office does.
Deterministic code handles
- Every numeric calculation
- Every claim-to-evidence anchoring
- Every compliance and integrity gate
- Routing, delivery, and the audit trail
Your Investment Office handles
- The six approval decisions
- Editorial control and tone of voice
- Configuration of segments and rules
- The final say on what ships
We will show you exactly where AI sits in the system, and exactly where it does not. The architecture is designed to be approved, not just demoed.
§ 03 - HOW IT WORKS
Five stages. Five points of control.
One continuous flow from raw data to delivered communication. Each stage has its own checkpoint; every checkpoint is auditable.
01 Candidate House View
02 IO Approval
03 Personalised Routing
04 Integrity Checks
05 Delivery + Audit Replay
01
Candidate House View
RiffSharp drafts a structured candidate - thesis, counter-arguments, balanced summary. Every claim is anchored to a source. Conviction stays within bounds you set. No model output is autonomous: it is a candidate, not a decision.
02
IO Approval
Your Investment Office reviews the candidate across six clear decisions: rating, conviction, evidence, counter-evidence, distribution, and framing for advisors and clients. Review depth is configurable - lighter gate-chain approval for RIAs; full editorial review for banks. Approval is what turns a candidate into a house view.
03
Personalised Routing
The approved view adapts to each reader along two dimensions: investor profile and DISC personality type - how each client prefers to receive information. Advisors get the deeper version of the same view: talking points, background, supporting evidence. The substance stays identical. Only register, tone, and emphasis change.
04
Integrity Checks
Every version is checked before it leaves. Rule-based gates are deterministic. A planned numeric-immutability check (CR-130) verifies byte-by-byte that no number was altered during personalisation. Hard launch blocker.
CR-130 - Planned - Launch Blocker05
Delivery + Audit Replay
Drafts flow into your CRM with placeholder slots. Voice-personalised advisor briefings carry advisor-only context, access-logged at every read. Every output is recorded in DecisionRecord - inputs, prompts, model versions, gate results, parameters, code version. The log is immutable, hash-chained, and replayable from your audit store independently of us.
§ 04 - PRODUCTS
One platform. Five communication surfaces.
Morning Edge
The approved house view - daily, weekly, monthly. Advisors receive the full briefing: talking points, context, evidence, counter-arguments. Each client receives the same view, in language tuned to their investor profile and DISC personality type.
Event Radar
Event-triggered: earnings, M&A, ratings actions, regulatory filings, market-moving news. Surfaced when they matter, routed only where they are relevant. Same approval workflow, same audit posture.
Risk Compass
Portfolio and macro-risk, translated into client-appropriate language. Risk-on / risk-off regime detection, liquidity and volatility signals - with full methodology disclosure for compliance and CIO review.
Signal Engine
The research and signal layer underneath: knowledge-graph anchoring, point-in-time data fabric, scanner harness, walk-forward backtesting. A configurable surface for firms that want deeper quantitative control.
Client OS
Outreach for life events, suitability shifts, and relationship triggers. Drafted for review - never auto-sent. Every touchpoint flows through the same approval and audit chain.
§ 05 - WHO IT IS FOR
One platform. Two segments. Two paths in.
Squeezed from above and below.
You sit between platform-scale aggregators above and wealthtech entrants below. Your moat is the relationship - depth, judgment, fiduciary fit. RiffSharp adds the communication layer that lets your existing team cover more clients, deeper, without losing what makes them yours.
Adoption is configuration, not migration. We sit above your CRM (Salesforce, Redtail, Wealthbox) and your custodian (Schwab, Fidelity, Pershing). Your data stays in your perimeter.
Personalisation at scale, without giving up control.
Your franchise rests on advisor relationships, institutional rigour, and compliance posture under BaFin, FINMA, and CSSF. The pressure is the same - next-gen HNW clients expect digital-native engagement; banker mobility is rising. RiffSharp scales personalisation without ever touching the approved view.
Pseudonymisation gateway, EU-resident endpoints (Frankfurt, Zurich), tenant-isolated configuration. Architecture review available with your CCO and Investment Office.
Every architectural decision was made with the compliance, model-risk, and audit teams who would have to approve the system as the first reader.
AI for prose. Code for substance.
Language models structure and draft. Deterministic code handles every audit-critical operation - every calculation, every claim anchor, every gate, every routing rule. No LLM ever performs an investment computation.
Firm-specific configuration
Per-firm configuration boundary. You set editorial tier, advisor voice, client segments, rating taxonomy, review rules, retention, and regional residency. Every component reads from your configuration - no shared defaults imposed on you.
Privacy-first client handling
Client identifiers are tokenised at your edge before processing. RiffSharp does not need names or account numbers to operate. Mapping stays with you. No independent path to re-identify clients.
Hash-chained audit trail
Every decision recorded - inputs, prompts, model versions, gate results, parameters, code version, firm configuration. Queryable from your audit store. Reproducible at any point in time. Hash-chain integrity verifiable independently of RiffSharp.
§ 07 - WHERE WE ARE
Early product. Serious architecture. Founding-client conversations.
Approach Built
- Knowledge graph of 870+ entities, point-in-time correct
- Entity-resolution benchmark: 88% auto-approve on held-out test set
- Quantitative feature library spanning macro, liquidity, and volatility regimes
- Six-pass review protocol and spec-conformance checks
Partners In Motion
- A Tier-1 European private bank - founding-client conversations
- A top-100 German wealth manager - introductory discussions
- Up to two founding-client engagements planned for this round
- Founding clients named as reference deployment for their segment
Before You Commit
- 90-minute architecture review with your CCO
- Methodology deep-dive with your quant team and Investment Office
- Pseudonymisation gateway walkthrough
- Reference architecture documentation, under NDA
"I have been on the receiving end of generic morning notes that arrive a day late and never reference my book. RiffSharp is the platform I wished my own bank had built."
Sebastian, founder. Private banking client himself.
§ 08 - TEAM
Built by people who understand regulated delivery.
Sebastian
Founder
Serial entrepreneur. Built and exited Dakoko, a Tier-2 automotive supplier, to a publicly listed acquirer. Risk discipline from signing personal guarantees. Quality obsession from shipping parts into moving vehicles. Private banking client himself - designed RiffSharp around the architecture he wished his own bank had built.
Nitin
Co-Founder - Architecture Lead
Owns the platform architecture and the deterministic engine - integration boundaries, fail-policy, the DecisionRecord schema, and the CR-130 numeric-immutability specification. Brings the technical delivery discipline regulated environments require.
Ivan
LLM & Knowledge Graph
Owns the Knowledge Graph (870+ entities, five-layer entity resolution), Event Radar, Entity Master, and the council subgraph. Internal entity-resolution benchmark: 88% auto-approve on held-out test set; 12% routed to manual review.
Kshitij
Quantitative & ML
Owns risk-on / risk-off regime detection, Risk Compass, the walk-forward backtest harness, and the quantitative feature library across macro, liquidity, and volatility regimes.
Amer
Senior QA & Verification
Medical-grade quality engineer. Owns the QA pipeline, the six-pass review protocol, the spec-conformance bot, and regression testing against frozen golden datasets. Verification at every layer.
Why this team.
Two founders built a company in automotive, where a quality failure can cost lives. That experience shaped every architectural decision. Deterministic gates - because we know what uncontrolled risk feels like, not from a textbook but from signing personal guarantees on a manufacturing line. Verification-first culture - because automotive teaches you that quality bolted on at the end fails in the field. Honest disclosure of what is not yet built - because over-claiming in regulated environments is a fatal credibility error.
- A SCENE Tuesday morning,
Tuesday morning,
twelve months from now.
7:42 AM. Earnings surprise from a top holding, overnight.
By 8:15, your Investment Office has approved the updated view.
By 8:45, every advisor has the briefing - talking points, evidence, counter-arguments - framed for their book.
By 9:00, every client is reading a note in their voice, on their terms.
Your CCO can replay any of it, down to the model version that drafted the prose.
§ 09 - FOUNDING CLIENT PROGRAM
Two seats. Direct influence. Reference deployment.
Founding clients help shape what gets built next.
Up to two founding-client engagements this round. Founding clients shape the configuration that ships in their tenant, become the named reference deployment for their segment, and inherit a configuration moat against subsequent customers.
What we offer before commitment
- 90-minute architecture review with your CCO and our architects - on-site or virtual
- Methodology deep-dive with your quant and Investment Office teams
- Pseudonymisation gateway live walkthrough
- Reference architecture documentation pre-commit, under NDA
- Master Claim Demo - fully synthetic illustrative artefact, end-to-end flow
What we ask of founding clients
- One Investment Office partner who joins monthly co-design sessions
- Configuration input across rating taxonomy, suitability classes, advisor voice profile, and consistency-check registry
- Reference call rights for subsequent qualified prospects
- A 90-day pilot scope on Morning Edge, ahead of broader advisor rollout
Request a meeting
A short note is enough. We respond within one business day. The first conversation is sixty minutes - a working discussion. No slides. No commitment.
Suggested next step Your Investment Office is the decision-maker.
Your Investment Office is the decision-maker.
We are the infrastructure that scales their judgment.
Book a working session - not a pitch. We will walk through the architecture, the compliance posture, the audit replay, and the founding-client structure. We will show you the Master Claim Demo. We will be straight about what is shipped, what is in progress, and what is planned.